Help with LED switcher supply

[technogeek]

Well then..... I guess there isn't two ways about it, I'm stuck into building two switchers.

 

[kchriste]

I'm stuck into building two switchers.

You can still do it with just one switcher. All you need is a low value resistor in each string to equalize the current between them. The resistors only have to drop a few volts so they shouldn't have to dissipate too much power. If you match the voltage drop of both LED strings as close as possible, you can use lower resistance and thus waste even less power as heat.

 

[technogeek]

You can still do it with just one switcher. All you need is a low value resistor in each string to equalize the current between them. The resistors only have to drop a few volts so they shouldn't have to dissipate too much power. If you match the voltage drop of both LED strings as close as possible, you can use lower resistance and thus waste even less power as heat.

I'm not interested in matching LED's for hours and hours....

So worst case scenario, the voltage drop difference between two strings could be 24V... which is 17W dissipation in the resistor...

 

[Sceadwian]

That's a lot of heat.

 

[technogeek]

exactly.

The current mirror would be a good solution if it never failed..... but the parts are cheap enough to build 2 switchers.. Just have to find the room to fit all the parts.

 

[Oznog]

Actually for a 700mA circuit, I would definitely prefer the HV9931 over the HV9910, because the 9931 has power factor correction. There is an electrical noise consideration here. But a 9910 will work ok.

Getting good heatsinking is a trick. There is critical factor here that you may not have noticed, and it might kill your project. In a Luxeon- and some of their clones- the thermal connection on back is NOT ELECTRICALLY NEUTRAL. It cannot be tied to either the + or - terminal either. So unless you can somehow make an electrically insulated, high performance thermal connection, each of the 40 devices must have its own heat sink and all of them must be isolated from either other and separated from human contact since it has high voltage on it.

I would drop the two-string idea competely, frankly. In the HV9910 you'd be using a 120VAC rectified, which may make up to 169V. Even with allowing a reasonable ripple on the caps making it down to 120vdc, that's still enough to drive 30x 4v devices. I think adding so much to the complexity to add like 10 more devices is unnecessary. Also notice the required inductance (and thus the total inductor size) is proportional to the difference between the input and output voltages. Making an LED string closer to the Vin is a good idea.

The circuit listed is not always stable. The current spike to turn on the transistor's gate will prematurely trip the current threshold under some circumstances. You need to put an RC filter on the current sense wire with a 300nS time constant to correct the problem.

 

[kchriste]

I'm not interested in matching LED's for hours and hours....

You probably won't find that much of a difference between LEDs if they are from the same batch. The 100 white LEDs I bought had a Vf spec of 2.75-3.80V but when I tested about 30 of them at the nominal current, they were all within 0.15V of one another.

I would drop the two-string idea competely, frankly.

If he really wanted to, he could use a voltage doubling circuit on the input to give him enough headroom for a single string. Two diodes and two caps is all that is needed for the voltage doubling / rectifier circuit. This is a common technique in computer power supplies (Switches between a doubler and a bridge circuit) to give you a 110/220Vac option via a switch.

 

[Oznog]

Matching Vf of the LEDs is not the whole problem. Two strings will go out of balance since the warmer string will develop a lower Vf, thus stealing more than half the current. The ballast resistor has to be large enough to keep this effect within a reasonable limit, but it will not eliminate it completely unless a large resistance is used which will be inefficient and make so much heat you may have to redesign the enclosure.

 

[kchriste]

Two strings will go out of balance since the warmer string will develop a lower Vf, thus stealing more than half the current.

That's true. He would have to physically arrange the LEDs on the heatsink (If he can isolate them electrically) so that one string is interleaved with the other so that, on average, the strings are at the same temperature. The best solution, as you pointed out, is a single string....

 

[justDIY]

I seem to recall seeing diagrams from lumileds or perhaps philips (for their xitanium driver) showing suggested wiring of series-parallel arrangement. The series strings were wired in parallel across the supply, and crossbars were used to link the two strings between each LED.

I supposed the purpose of the crossbar was to normalize voltage on the two strings, preventing any one abnormal led from causing a current hogging scenario.

I'm searching for that info again, will post the link when I find it.

edit: looks like LuxeonStar.com summarized the application note here:
**broken link removed**

 

[audioguru]

Guess who bought Lumileds and makes Luxion bright LEDs?

 

[justDIY]

Philips co-founded lumileds with HP (aka Agilent). Only recently did Philips buy-out HP's stake in the company, making lumileds wholey owned by Philips

edit: typo!

 

[audioguru]

I have some very nice HP and Agilent LEDs. I didn't know Philips were involved with them.

 

[technogeek]

I seem to recall seeing diagrams from lumileds or perhaps philips (for their xitanium driver) showing suggested wiring of series-parallel arrangement. The series strings were wired in parallel across the supply, and crossbars were used to link the two strings between each LED.

I supposed the purpose of the crossbar was to normalize voltage on the two strings, preventing any one abnormal led from causing a current hogging scenario.

I'm searching for that info again, will post the link when I find it.

edit: looks like LuxeonStar.com summarized the application note here:
**broken link removed**

Wow! Now that's strange!

So if I were to create a 1.4A power supply, and used two strings of 20... what are the chances each would get exactly 700mA? Aaaand, how many crossover connections should there be? One for every 5 LED's?

Looking at their datasheet- https://www.luxeonstar.com/xitanium-120v-drivers.pdf

they seem to prefer low voltage high current outputs, with 6 LED's per branch..... Interesting.

edit- seeing as how I already ordered parts, I plugged the numbers in again for double the current. The only things that changed were the input capacitor trippled in size, and the inductor halved. All of the currents/voltages are still well within the limits of the parts I chose. So, what would happen if I used a 1.8mH inductor in place of a 0.9mH inductor? Anything? Less current ripple maybe?

Oh, and my sense resistor halved.

 

[justDIY]

using a larger inductor offers no benefit, and as far as I know, no major drawbacks. However, there is a drawback that the larger inductor may not be rated for your newly calculated peak current.

 

[Oznog]

I prefer the Lamina devices myself. Outrageously flat and an electrically insulated backing of pretty decent size.

I noticed that the 10mm LEDs have evolved into ones with HUGE ratings. I ordered some off a guy on eBay with a 150,000mcd rating over 30 degress! I mean, holy crap! They use multiple dies in the package, which precludes a tight focus. I'm still waiting, they came from China so it's expected to take awhile. I only paid $5.62 for 5 of them though.

 

[technogeek]

I prefer the Lamina devices myself. Outrageously flat and an electrically insulated backing of pretty decent size.

I noticed that the 10mm LEDs have evolved into ones with HUGE ratings. I ordered some off a guy on eBay with a 150,000mcd rating over 30 degress! I mean, holy crap! They use multiple dies in the package, which precludes a tight focus. I'm still waiting, they came from China so it's expected to take awhile. I only paid $5.62 for 5 of them though.

What wavelength? 32 lumens is about the equivalent of a 1W or 3W LED depending on wavelength...

 

[justDIY]

the candella is a worthless unit (and method) of "measurement"

depend on lumens or mw or some other more reputable method